Cyclist detection in LIDAR scans using faster R-CNN and synthetic depth images

Saleh, Khaled J.; Hossny, Mohammed; Hossny, Ahmed; Nahavandi, Saeid

doi:10.1109/itsc.2017.8317599

Cited by 34 publications

(18 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This means landmarks could be detected but may not be precise or have high accuracy. Comparing to well-configured examples presented in [19][20][21], the accuracy of our Faster R-CNN was moderately lower. This reduction in detection accuracy is the cause of a lower localization accuracy, as landmark detection results are required to generate localization results in the Faster R-CNN localization method.…”

Section: Detection Experimentsmentioning

confidence: 77%

“…The increased speed of Faster R-CNN for object detection makes it suitable for real time applications [17,18]. Implementations of Faster R-CNN spread throughout various applications, such as the detection of cyclists in depth images [19], pedestrian detection from security cameras [20], and ship detection in remote sensing images that contain foggy scenes [21]. In [19][20][21] it is shown that Faster R-CNN has high accuracy (more than 80%), slightly higher than human volunteers that have approximately 75% accuracy [22].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning-Based Landmark Detection for Mobile Robot Outdoor Localization

Nilwong

Hossain²,

Kaneko

et al. 2019

Machines

View full text Add to dashboard Cite

Outdoor mobile robot applications generally implement Global Positioning Systems (GPS) for localization tasks. However, GPS accuracy in outdoor localization has less accuracy in different environmental conditions. This paper presents two outdoor localization methods based on deep learning and landmark detection. The first localization method is based on the Faster Regional-Convolutional Neural Network (Faster R-CNN) landmark detection in the captured image. Then, a feedforward neural network (FFNN) is trained to determine robot location coordinates and compass orientation from detected landmarks. The second localization employs a single convolutional neural network (CNN) to determine location and compass orientation from the whole image. The dataset consists of images, geolocation data and labeled bounding boxes to train and test two proposed localization methods. Results are illustrated with absolute errors from the comparisons between localization results and reference geolocation data in the dataset. The experimental results pointed both presented localization methods to be promising alternatives to GPS for outdoor localization.

show abstract

Section: Detection Experimentsmentioning

confidence: 77%

Section: Introductionmentioning

confidence: 99%

Deep Learning-Based Landmark Detection for Mobile Robot Outdoor Localization

Nilwong

Hossain²,

Kaneko

et al. 2019

Machines

View full text Add to dashboard Cite

show abstract

“…However, these techniques require a handful amount of labelled data for training them which is both time-consuming and cumbersome to get for many tasks. Thus, the utilisation of synthetic data for training such techniques got some momentum over the past few years [7], [8]. With synthetic data, the process for obtaining groundtruth labels becomes much easier and automated most of the time.…”

Section: Introductionmentioning

confidence: 99%

Domain Adaptation for Vehicle Detection from Bird's Eye View LiDAR Point Cloud Data

Saleh

Abobakr

Attia

et al. 2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

Self Cite

View full text Add to dashboard Cite

Point cloud data from 3D LiDAR sensors are one of the most crucial sensor modalities for versatile safety-critical applications such as self-driving vehicles. Since the annotations of point cloud data is an expensive and time-consuming process, therefore recently the utilisation of simulated environments and 3D LiDAR sensors for this task started to get some popularity. With simulated sensors and environments, the process for obtaining an annotated synthetic point cloud data became much easier. However, the generated synthetic point cloud data are still missing the artefacts usually exist in point cloud data from real 3D LiDAR sensors. As a result, the performance of the trained models on this data for perception tasks when tested on real point cloud data is degraded due to the domain shift between simulated and real environments. Thus, in this work, we are proposing a domain adaptation framework for bridging this gap between synthetic and real point cloud data. Our proposed framework is based on the deep cycle-consistent generative adversarial networks (CycleGAN) architecture. We have evaluated the performance of our proposed framework on the task of vehicle detection from a bird's eye view (BEV) point cloud images coming from real 3D LiDAR sensors. The framework has shown competitive results with an improvement of more than 7% in average precision score over other baseline approaches when tested on real BEV point cloud images.

show abstract

“…As a representative, convolutional neural network (CNN) has shown extraordinary application value in the field of image recognition. As one of the important CNN models, Faster R-CNN has achieved many exhilarating results, such as pedestrian detection [11]- [14], vehicle detection [15], [16], sign recognition [17], [19], medical detection [20], biological detection [21], remote sensing [22], and other image detections [23]- [25]. The main development direction is to improve the recognition rate and expand the application range by modifying the network structure.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning-Based Interference Fringes Detection Using Convolutional Neural Network

Zhang

Song

et al. 2019

IEEE Photonics J.

View full text Add to dashboard Cite

The interference fringes of interferometry are the key to reconstruct a threedimensional topography. But currently the adjustment of the fringes is done by manual, which is time-consuming and lack of quantitative control. Due to the complexity of the fringes, the traditional methods have low recognition rates and are only suitable for the ideal fringes in specific cases. Therefore, an interference fringes discovery (IFD) model consisting of "fringes region proposal network" (FRPN) and "fringes stitching" (FS) model is proposed. The FRPN, a deep convolutional neural network modified on Faster R-CNN, accurately recognizes the fringes with identification boxes. By integrating the feature maps of multiple layers and fine-tuning, the ability of the modified network to extract more complicated fringes is improved. Based on the identification boxes generated by FRPN, the FS restores the fringe shape and generates a complete recognition area. The IFD model achieves 98% accuracy on our testing set. The experimental results confirm that our model has the excellent performance on fringes-detection. It provides important support for the automation and the precise measurement of interferometers.

show abstract

Cyclist detection in LIDAR scans using faster R-CNN and synthetic depth images

Cited by 34 publications

References 17 publications

Deep Learning-Based Landmark Detection for Mobile Robot Outdoor Localization

Deep Learning-Based Landmark Detection for Mobile Robot Outdoor Localization

Domain Adaptation for Vehicle Detection from Bird's Eye View LiDAR Point Cloud Data

Deep Learning-Based Interference Fringes Detection Using Convolutional Neural Network

Contact Info

Product

Resources

About